Data Sourcing. InfoTrie’s algorithms scan thousands of websites, blogs, and business news publications every 5 minutes, searching for company mentions. Each article that mentions a given company is assigned a “relevance” score with respect to that company. Articles with low relevance scores are discarded; only articles with moderate and high relevance scores are kept for analysis.
Once a corpus of highly relevant articles has been selected, InfoTrie applies natural language processing in a machine learning framework, to assess the sentiment associated with that selection. Aggregate volume and sentiment scores are then computed using the individual article scores and their relevance weights.